VoltDB and Custom Search
Feb 26, 2014
Hello,
I understand the way voltdb is architectured and what implication it brings with it. We are currently need to implement a search that is using a geo search including an attribute selection. The idea is to store an array of sorted attributes per user and select while searching for the user.
Since this involves somewhat 10k+ entries after index access it will be easy. The array will be simply a blob that can be accessed bytewise to extract the integer values(?). Each array might be around 100bytes or less. So we accessing just 1MB of data during the search.
Since all those stored procedures are executed in parallel, we would just add more replicas to the problem and done.
My question targets the insertion vs the search problem. We try to use the VoltDB for quite a large high volume site where there are maybe a 1000 changes per second are necessary to track the users (geowise).
VoltDB should drive a complete social application website + mobile which we managed to use isolated tables with not much of interconnection (no foreign keys mostly since users can not be removed and either languages cant).
Is there any problem with that? Is this to much load.
When I do a database query with the sql part of a stored procedure do I access the actual memory object or do I just get a copy (important for the memory footprint etc).
I understand the way voltdb is architectured and what implication it brings with it. We are currently need to implement a search that is using a geo search including an attribute selection. The idea is to store an array of sorted attributes per user and select while searching for the user.
Since this involves somewhat 10k+ entries after index access it will be easy. The array will be simply a blob that can be accessed bytewise to extract the integer values(?). Each array might be around 100bytes or less. So we accessing just 1MB of data during the search.
Since all those stored procedures are executed in parallel, we would just add more replicas to the problem and done.
My question targets the insertion vs the search problem. We try to use the VoltDB for quite a large high volume site where there are maybe a 1000 changes per second are necessary to track the users (geowise).
VoltDB should drive a complete social application website + mobile which we managed to use isolated tables with not much of interconnection (no foreign keys mostly since users can not be removed and either languages cant).
Is there any problem with that? Is this to much load.
When I do a database query with the sql part of a stored procedure do I access the actual memory object or do I just get a copy (important for the memory footprint etc).